Model guide · current service distinction
Stable Diffusion 3.5 Large Turbo
Large Turbo is Stability AI's distilled SD 3.5 variant designed for a small number of inference steps. This guide explains that design and where it differs from the image generator available here today.
What 'Turbo' means here
Stability AI describes SD 3.5 Large Turbo as a distilled version of Large designed to produce an image in four inference steps. The step count refers to that model's inference design, not a guarantee that a browser request finishes in a fixed number of seconds.
Real latency depends on queueing, service capacity, resolution, network conditions, and setup. Quality also varies by prompt and subject. Consult the model card for actual deployment instructions and limitations rather than transferring a model claim to this service.
Read the official model card ↗Write for a quick visual decision
For fast iteration, make the subject, placement, lighting, and medium unambiguous. Change one visual decision at a time and compare outputs. This method helps whether another tool offers Turbo or you are using this site's current generator; it does not make the two models equivalent.
Example direction
A compact reading nook by a tall window, one rust-colored armchair on the left, warm late-afternoon side light, pale plaster walls, quiet editorial interior photograph.
Check geometry, text, anatomy, and any real-world details rather than treating speed-oriented output as publication-ready by default.
Speed is a workflow property
Large Turbo's reduced inference-step design may help a compatible deployment iterate faster than Large. It does not establish this site's generation latency. Our own requests may queue, fail, or vary in duration.
If an SD 3.5 comparison is essential, choose a tool that exposes the exact Large and Turbo checkpoints, then compare the same prompt and settings. This site currently offers neither checkpoint as a model choice.
License and output rights
Stability AI's Community License governs this model and contains conditions for use and distribution. It does not automatically clear third-party rights in an input or output. Our site's Terms govern images made through our current tools; inspect source-image permissions, recognizable people, brands, and similar protected work before commercial use.